AcTiVComp20

Competition on Superimposed Text Detection and Recognition in Arabic News Video Frames

hold within the framework of the 25th International Conference on Pattern Recognition (ICPR2020),Milano- Italy 13 -18 September 2020

Call For Participation (PDF)


Due to the escalated situation of the COVID19 and the difficulties that it creates to researchers, we have moved the submission deadline to August 7th, 2020

Background and relevance to ICPR community

Among the pattern recognition fields, automatic text recognition, known as OCR, has been widely studied for its prominent position in our everyday life. OCR has a long history of research that started from isolated character recognition and evolved to printed/handwriting document recognition. Recently, embedded texts in videos and natural scenes have received increasing attention as they often give crucial information about the media content. Nevertheless, extracting text from such content is a non-trivial task due to many challenges like the variability of text patterns. Over the last decade, interest in this area of research has led to a plethora of text detection and recognition methods. So far, these methods have focused only on some languages such as Latin and Chinese [1]. For a language like Arabic, which is used by more than one billion people around the world, the literature is limited to very few studies, and most of the existing methods have been tested on private datasets with non-uniform evaluation protocols, which makes direct comparison and scientific benchmarking rather impractical. The present competition aims to fill the aforementioned gap by encouraging Scene/Video-OCR researchers to develop and test their systems on a standard annotated dataset and using the same evaluation protocols/metrics.


Video-OCR system pipline
Video-OCR: context, challenges and objectives

Motivation and objectives

The suggested competition represents a part of the AcTiVComp series that have been organized respectively within the ICPR'16 [2] and ICDAR'17 [3] conferences, using the AcTiV dataset [4] as a benchmark. Actually, the former editions have attracted seven groups for participating and have received ten systems in total. The best achieved F-score for the channel free detection protocol was 0.85. For the recognition task, the best results in the channel free protocol have not exceeded 0.76 in terms of Line Recognition Rate (LRR). Furthermore, the obtained results on the recently added subset (SD 480x360) were quite low for almost all the participating systems. For this reason, we believe that there is a requirement to organize a new edition in order to improve the detection and recognition results, especially for the channel free evaluation protocols. Additionally, this third edition has a special focus on the evaluation of end-to-end Arabic Video-OCR systems where, to our knowledge, very little work has been realized till now; i.e. most of the existing systems worked solely in text detection or text recognition.

How to participate

Important Dates

March 04, 2020: Registration open
March 07, 2020: Training set available
July 01, 2020 July 27, 2020: Test set available
July 18, 2020: August 7, 2020 Submission of final results and executables / source codes
July 19, 2020: August 8, 2020 Submission of a short description of the participating method(s)


NOTE: executable / source code submission is highly recommended but not mandatory for participating to the competition.


References
[1] https://www.springer.com/gp/book/9781447165149
[2] O. Zayene, N. Hajjej, S. M. Touj, S. Ben Mansour, J. Hennebert, R. Ingold et N. E. Ben Amara. “ICPR2016 Contest on Arabic Text Detection and Recognition in Video Frames —AcTiVComp”. In 23rd International Conference on Pattern Recognition (ICPR), p. 187-191, Decembre 2016.
[3] O. Zayene, J. Hennebert, R. Ingold et N. E. Ben Amara. “ICDAR2017 Competition on Arabic Text Detection and Recognition in Multi-resolution Video Frames”. In 14th International Conference on Document Analysis and Recognition (ICDAR), p. 1460-1465, Novembre 2017.
[4] O. Zayene, J. Hennebert, S. M. Touj, R. Ingold, and N. Essoukri Ben Amara, “A dataset for Arabic text detection, tracking and recognition in news videos- AcTiV”, in Proc. of the 13th International Conference on Document Analysis and Recognition (ICDAR'15), Nancy- France, August 2015.